82% assignment of protons is VERY low. Extreme caution would be advisable before proceeding to the restraint generation stage before completion of proton assignments.

For proteins, many observable peaks in the NOESY correspond to intra-residue and sequential-residues and may not be of much use for tertiary structure determination. Since the crystal structure is available, you may check if you are limited by sensitivity by identifying the longest distance in the crystal structure for non-sequential NOe's involving backbone protons observable in your spectra. If this value is significantly lower than 0.45 nm, then you may be able to obtain more restraints by increasing the S/N of your spectra. The S/N ratio for your spectra may be increased by

• increase of mixing time (Caution !!),
• increase of protein concentration,
• increase of field strength (preferable),
• use of cryoprobe/10 mm probe (instead of the standard 5mm RT),
• change of temperature (optimum temperature of NOESY is usually not always the same as the optimum temperature for 1D or HSQC spectra),

• change of pH, etc.

  In addition, proper choice of the pulse sequences and optimization of the delays to match the observed relaxation times is often critical for obtaining good quality X-edited NOESY spectra.

2D-NOESY (homonuclear) spectra often contain many non-sequential NOe peaks that are not observable in other NOESY spectra, most of these are usable as restraints if a low resolution structure is used as a filter to assist in the assignment process.

Some proteins may adopt a specific conformation in the solid (crystal), but may have substantial local flexibility in solution. in such cases, especially for helical proteins, it may be difficult to observe many non-sequential NOe's. Restraints from other types of data, such as J-coupling, are less sensitive to low amplitude local dynamical motion.